Building An Auto-Scaling Cluster¶

The senlin service provides a rich set of facilities for building an auto-scaling solution:

• Operations: The CLUSTER_SCALE_OUT, CLUSTER_SCALE_IN operations are the simplest form of commands to scale a cluster. The CLUSTER_RESIZE operation, on the other hand, provides more options for controlling the detailed cluster scaling behavior. These operations can be performed with and without policies attached to a cluster.
• Policies: The senlin.policy.scaling (link) policy can be applied to fine tune the cluster scaling operations. The senlin.policy.deletion (link) policy can be attached to a cluster to control how nodes are removed from a cluster. The senlin.policy.affinity (link) policy can be used to control how node affinity or anti-affinity can be enforced. The senlin.policy.region_placement (link) can be applied to scale a cluster across multiple regions. The senlin.policy.zone_placement (link) can be enforced to achieve a cross-availability-zone node distribution.
• Receivers: The receiver (link) concept provides a channel to which you can send signals or alarms from an external monitoring software or service so that scaling operations can be automated.

This section provides some guides on integrating senlin with other services so that cluster scaling can be automated.